Wan 2 7

This wraps Wan-AI's flagship text-to-video model through RunComfy's CLI, letting you generate 2-15 second clips at 720p or 1080p with audio-driven lip sync if you supply a track. The main draw is the audio_url parameter for syncing mouth movement to your own voiceover, plus multi-reference conditioning and solid motion physics. You get prompt expansion by default, which rewrites short prompts for better results, but you can disable it for literal control. Best for lip-synced ads or multi-language dub variants where you need the same visual with different audio. Duration caps at 15 seconds and there's no 4K, so plan to stitch longer narratives from multiple calls.